A typical plasma etching apparatus comprises a reactor in which there is a chamber through which reactive gas or gases flow. Within the chamber, the gases are ionized into a plasma, typically by radio frequency energy. The highly reactive ions of the plasma are able to react with material, such as a polymer mask on a surface of a semiconductor wafer being processed into integrated circuits (IC's). Prior to etching, the wafer is placed into the chamber and held in proper position by a chuck or holder which exposes a top surface of the wafer to the plasma. There are several types of chucks (also sometimes called susceptors) known in the art. The chuck provides an isothermal surface and serves as a heat sink for the wafer. In one type, a semiconductor wafer is held in place for etching by mechanical clamping means. In another type of chuck, a semiconductor wafer is held in place by electrostatic force generated by an electric field between the chuck and wafer. The present invention is applicable to both types of chucks.
In a typical plasma etching operation, the reactive ions of the plasma chemically react with portions of material on a face of the semiconductor wafer. Some processes cause some degree of heating of the wafer, but most of the heating is caused by the plasma. The chemical reaction rate between the materials in the plasma and the wafer material, on the other hand, is accelerated to some degree by the temperature rise of the wafer. Local wafer temperature and rate of chemical reaction at each microscopic point on the wafer are related to an extent that harmful unevenness in etching of material over a face of the wafer can easily result if the temperature of the wafer across its area varies too much. In most cases, it is highly desirable that etching be uniform to a nearly perfect degree since otherwise the Integrated Circuit devices (ICs) being fabricated will have electronic characteristics that deviate from the norm more than is desirable. Furthermore, with each increase in the size of wafer diameter, the problem of ensuring uniformity of each batch of ICs from larger and larger wafers becomes more difficult. In some other cases, it would be desirable to be able to control the surface temperature of the wafer to obtain a custom profile.
The problem of temperature rise of a wafer during reactive ion etching (RIE) is well known, and various attempts in the past to control the temperature of a wafer during RIE have been tried. FIG. 1 illustrates one way to control wafer temperature during RIE. An inert coolant gas (such as helium or argon) is admitted at a single pressure within a single thin space 102 between the bottom of the wafer 104 and the top of the chuck 106 which holds the wafer 104. This approach is referred to as backside gas cooling.
There is generally no o-ring or other edge seal at the chuck perimeter except for a smooth sealing land extending from about 1 to 5 mm at the outer edge of the chuck 106 in order to reduce coolant leakage. Inevitably, without any elastomer seal, there is significant and progressive pressure loss across the sealing land, such that the edge of the wafer 104 is inadequately cooled. The heat generated near the edge of the wafer 104 must therefore flow significantly radially inward before it can effectively be conducted away to the chuck. The arrows 106 on top of the wafer 104 illustrate the incoming flux heating the wafer 104. The flow of the heat in the wafer 104 is illustrated with the arrows 110. This explains why the edge zone of the chuck always tends to be hotter than the rest of the surface. FIG. 2 illustrates a typical temperature distribution on the wafer 104. The pressure loss at the peripheral portions of the wafer 104 causes the wafer 104 to be much hotter at the peripheral portions.
One way of dealing with the need for zone cooling is to vary the surface roughness or to cut a relief pattern to effectively change the local contact area. Such a scheme can be used without backside coolant gas at all, in which case the contact area, surface roughness, and clamp force determine the heat transfer. However the local contact area can only be adjusted by re-machining the chuck. Another way of dealing with the need for zone cooling is to use coolant gas whose pressure is varied to increase and fine tune thermal transport. However the relief pattern is still substantially fixed. By dividing the surface of the chuck into different zones, with or without small sealing lands as dividers, and supplying separate cooling gasses to each zone, a greater degree of independent spatial control may be achieved. The gas supply to each zone may have different composition or be set to a different pressure, thus varying the thermal conduction. Each zone's operating conditions may be set under recipe control, or even dynamically stabilized during each process step. Such schemes depend on redistributing the incoming heat flux from the plasma and driving it into different regions. This is relatively effective at high power flux but will only give small temperature differentials at lower power flux. For instance, with about 3 to 6 W per cm.sup.2 of uniform flux and about 3 mm sealing land, it is possible to get center to edge thermal gradients that lead to a 10° C. to 30° C. temperature increase near the wafer periphery. Thermal gradients of this magnitude can be very effective as a process control parameter. For example, radial variations in plasma density or asymmetries in a reactor layout, which can affect critical process performance metrics, are counteracted with an appropriate substrate temperature pattern. However, some processes may run at low power, for instance poly gate processes, may have a flux of only 0.2 W per cm.sup.2. Unless the average conduction is made extremely low, which is very difficult to control and tends to result in inadequate overall cooling, then there will be only a very small differential of typically less than 5° C.
Accordingly, a need exists for a method and, apparatus for controlling the temperature of semiconductor wafers during reactive ion etching and similar processes without requiring significant plasma heat flux. A primary purpose of the present invention is to solve these needs and provide further, related advantages.